I’m tracking down the world’s best community builders. Here’s what I’ve learned so far...

In 2021, Paz Pisarski asked her network whether anyone else who builds communities felt this alone, and fifty people answered. They ate sushi in a deserted coworking space and found the loneliness was almost unanimous.

That meetup became the Community Collective, which now has 700 members across 18 countries and a waitlist of nearly a thousand...

I’ve spent the last few episodes of the Discourse Podcast talking to some of the folks at the coalface of community in 2026: Paz, Sam Saffron and Sarah Hawk (co-CEOs of Discourse), and Richard Millington, who has been building communities since he was a teenager, working with Apple, Google, Meta, and about three hundred other organisations.

I wanted to put down what I’ve learned so far: what’s working, what’s not, and where this is going.

(Remember! If you have a guest you’d love to see us chat with, drop us a line at marketing@discourse.org)

Belonging is the product, skills are the byproduct

Paz splits everything she builds into two parts. The hardcore curriculum is what people sign up for: public speaking, strategy, operations. The hidden curriculum is why they remain: the rituals, the check-ins, the matched-up cohort partners, the orange everyone wears. People come for the skills and stay for the feeling that they made the best decision of their lives.

Richard started out, in his own words, a “purist,” convinced a real community meant deep, name-everyone, part-of-something belonging. His take today is more permissive: if people are helping each other and getting something out of it, that’s a community, even if it’s a Facebook page where someone occasionally comments on a beauty brand.

You build it with people, not for them

Paz calls herself a community architect - a label I loved. A good architect doesn’t simply design your dream house and then hand you the keys; instead, they consult you at every step, get feedback, and co-design with you.

You can't build community for people; you have to build it with them. Set the parameters, offer the options, let them choose.

In fact, when Paz' in-person meetup needed to go online, she emailed everyone two platform options and let them vote - and someone even offered a paid Zoom account.

Most communities don’t crash. They never take off

I asked Richard what separates the communities that work from the ones that crash and burn. Communities rarely collapse, he said. They just never get off the ground...

The communities that thrive have a founder who started it because they cared, grew it slowly, and was already part of the space they were serving. The ones that fail are usually a games publisher or a brand trying to brute-force thousands of members into a room and shout “be a community now.”

The measurement problem is real, and the answer is letting go

Every conversation came back to ROI, and everyone admitted it’s brutally hard. Richard pointed me to Jerry Z. Muller’s The Tyranny of Metrics, which argues that once you set a measurable goal, people optimise for it at the expense of everything else. His favourite example: a community manager measured on engagement boosted her numbers by removing the spam filters, hit her target, and destroyed the thing she was measuring - Goodhart’s law in action.

The community architect has a harder job than the sales manager, as Paz pointed out, because she’s serving two stakeholders at once, the business and the members, and the members’ needs shift week to week.

Richard’s fix is to stop chasing precision: ask a net-promoter-style question instead, something like “to what extent did the community influence your decision to renew?”, and treat the answer as good enough.

The walls are coming down, whether you like it or not

Hawk is haunted by the communities she used to run that no longer exist...

She thinks of community builders as stewards of the knowledge inside them, which makes the platform question existential rather than operational.

Her advice is to design your exit strategy before you start. If you’re on a centralised platform, the community exists at a centralised decision maker's discretion. On a proprietary platform, as someone once told her, you’re a hostage rather than a customer. Open source is the only place where the code survives no matter what happens to the company that hosts it.

On AI, Hawk and Richard agree: don’t fight the trend. The easiest way to protect a community’s knowledge is to spread it widely: clean up your titles, fix your tags, let the models read you well. Richard compares it to the retail apocalypse, where some stores survive, most go through a brutal sorting, and the winners shift from platforms to programs.

AI should never pretend to be human

At Discourse, AI content always gets labelled. No robot is allowed to pose as a person. The fastest way to empty a community is to let members discover that half of them were bots, and the bots are getting better at talking.

The useful AI, in Sam’s view, is the boring kind. Spam detection. Translation. Tagging, which humans are terrible at and machines are weirdly good at. Crunching a week of conversation into a report for an overwhelmed moderator who can’t read all of it. The machine flags suspicious content, and a person decides what to do with it.

The mistake companies make is bolting AI on as a tagline. If the only problem you’re solving is “we don’t have AI yet,” it’s likely the wrong problem.

What I’m sitting with

I started these conversations thinking community was something you market, measure, and optimise. I’m not so sure anymore. That’s what I’m hoping to dig into this season of the Discourse podcast, and with more episodes coming, maybe I’ll find some answers...

The people best at this keep telling me to let go of the idea that you get to fully control the room. Hawk compared communities to rock bands: the fact that a band was great for a few years and then broke up doesn’t mean it shouldn’t have existed. There’s something wonderful about a group of humans gathering in one spot, physical or digital, and giving a damn together...

Paz’s assignment is small, achievable, and direct: if you want to build a community, find one person this week, a stranger or a friend, and share something you learned that actually helped you.

That’s how it starts.